Isotopic tracers are useful in provenance and mobility applications across many different disciplines (Bentley, 2006; Crowley et al., 2015). The ratio of 87Sr/86Sr has proven invaluable to these studies since it reflects geological context and does not exhibit fractionation across trophic levels (Hoppe et al., 1999; Price et al., 2002). In addition, because the Rb to Sr ratio is so low and the time involved in studies so short (typically on Holocene or Late Pleistocene materials), the 87Sr/86Sr ratio does not evolve by radioactive decay and can be treated as essentially a stable isotope (Bowen and West, 2008). However, the impact of these techniques is highly dependent on the quality of baseline comparative data (i.e., isoscapes), complicating its use in areas dominated by allochthonous surface sediments of alluvial, eolian, or glacial origin.
The North American Midcontinent is one of these areas (Figure 1). Extensive glaciation during the Late Pleistocene and the subsequent geomorphic history of the region means that bedrock is often deeply buried and contributes little to surface soil parent material. Surface deposits are a heterogeneous mix of eolian, alluvial, and glaciogenic origin. Although there have been previous attempts to model regional 87Sr/86Sr composition in the region (Beard and Johnson, 2000; Bataille and Bowen, 2012), the resolution of surface geological mapping and the wide variability in surface geomorphology, even within small areas, have complicated efforts to resolve a Sr isoscape for surface materials. Although direct surface sampling of plants, animals, and sediments (Price et al., 2007; Hedman et al., 2009; Widga et al., 2010; Slater et al., 2014) offers some insight into regional Sr distribution patterns, these efforts have been limited to small parts of the study area and are opportunistic surveys undertaken independent of local geomorphic information.
Variability in surface 87Sr/86Sr is controlled by geological processes operating at two levels. First-order variability is determined by the source rock of surface deposits. Generally, 87Sr/86Sr of glaciogenic deposits is determined by the provenance of glaciogenic materials and subsequent landscape evolution (Taylor and Faure 1981, Widga et al., 2010), whereas 87Sr/86Sr of pre-glacial bedrock sources may be influenced by the age of parent rocks (Beard and Johnson, 2000), degree of crustal recycling (Bataille et al., 2014), progressive evolution of marine Sr reservoirs (McArthur et al., 1994), regional volcanism (Leeman 1982), or a number of other processes. Second-order variability is determined by local-scale geomorphological processes occurring within a geologic unit (Carpenter et al., 2003, Faure 1986, Widga et al., 2010).
Strontium is a common element found in geological, hydrological, and ecological systems at the Earth’s surface. It has four, naturally occurring stable isotopes (Faure, 1986): 84Sr, 86Sr, 81Sr, and 88Sr. The isotope 87Sr, however, is the radiogenic daughter isotope of 87Rb which has a half-life of 48.8 × 109 yr. The ratio of this decay product (87Sr) to the stable isotope 86Sr increases as the age of a rock formation increases. Due to its relatively large atomic mass, 87Sr/86Sr changes little as it passes from weathered bedrock, to soil parent material, and ultimately to plants and primary consumers (Flockhart et al., 2015). This ratio is dependent on the age and Sr content in soil parent material and atmospheric inputs (Beard and Johnson, 2000; Bataille et al., 2012; Bataille and Bowen, 2012; Reynolds et al., 2012). Geomorphic processes such as weathering and transport of parent rocks complicates surface 87Sr/86Sr distribution in many parts of the world where allochthonous sediments comprise soil parent material.
Within the Midcontinental study area, 87Sr/86Sr values are determined by the sediment source(s), local weathering processes, and surface geomorphology. The region is drained by a number of large rivers. Valley alluvium reflects a variety of local and non-local sources, including upstream deposits. Major basins in the study area include the Mississippi, Illinois, Ohio, Wabash, Missouri, Platte, Kansas, and Arkansas River basins. In the Great Plains, these rivers begin in the foothills of the Rocky Mountains and carry a sediment load reflecting these origins. For instance, the upper Missouri River drains extensive Neogene deposits in Montana and western North Dakota. The potential contribution of these young rocks on downstream alluvial deposits is evident in the lower Missouri valley 87Sr/86Sr in Nebraska and Iowa (Widga et al., 2010). Alternatively, erosion of local shale or carbonate units may contribute significantly to valley Sr ratios. In eastern parts of the study area, terrace fill is commonly derived from glaciogenic sediments such as reworked glacial till or loess deposits.
Loess covers a large portion of the study area and can reach depths of >40 m (Bettis et al., 2003). The source for loess deposits is in nearby river valleys, although long-distance transport of clay and silt-sized particles can occur (Muhs and Bettis, 2000; Muhs et al., 2013). The prevailing wind direction for most of the study area is from the northwest. Therefore the source area for loess deposits in the Great Plains are typically from non-glaciogenic sources (Mason, 2001) while loess in areas covered by recent glacial deposits reflects a glacial origin via large river valleys (Muhs and Bettis, 2000). Both regions may be susceptible to long-distance transport of dust (Mason and Jacobs, 1998).
Spatial variation in surface Sr isotope ratios provide the basis for many recent studies of human and animal mobility, as well as source estimates for sedimentary deposits. Small mammals and animals with limited mobility are typically used to assess the range of local variation in archaeological mortuary populations (Price et al., 2007; Hedman et al., 2009; Slater et al., 2014). This study however, relies on the expression of surface 87Sr/86Sr in regional grasses, an approach previously used to infer mobility middle Holocene bison populations from the eastern Great Plains (Widga et al., 2010).
Materials and methods
Sampling occurred opportunistically over the course of more than a decade. The distribution of samples reflects large-scale differences in regional landscape evolution, such as differences in loess deposition (Bettis et al., 2003), the evolution of alluvial systems (Mandel, 2008), and glaciogenic landscapes (Ehlers and Gibbard, 2004). Grass samples were collected from lithologically defined locations along roadsides or within natural areas. Locations were selected to avoid the effects of agricultural runoff, non-local rock products (e.g., roads surfaced with crushed limestone), or modern earth moving activities. Grasses are an ideal material to assess levels of bioavailable Sr because 1) they form the base of regional terrestrial foodwebs and incorporate 87Sr/86Sr in similar proportion to the soil parent material (Capo et al., 1998); 2) they are minimally impacted by atmospheric inputs of Sr due to low leaf area; and 3) they are widespread and common elements of the regional flora.
The emergent portion of grasses was collected and rinsed in DI water. Dried grass samples were then ashed at 425°C in platinum crucibles. The ashed grasses were dissolved in 3.5 N HNO3 and eluted through Eichrom Sr-spec ion-exchange columns. The purified Sr was then analyzed for 87Sr/86Sr ratios on a Thermal Ionization Mass Spectrometer (TIMS), an automated VG Sector or VG Sector 54 in the University of Kansas Isotope Geochemistry Laboratories. Isotope ratios were adjusted to correspond to a value of 0.71250 on NBS-987 for 87Sr/86Sr. Over 120 runs of NBS 987 were made along with the samples. Standard deviation of results is about 11ppm on ratio, and standard error is miniscule. We also assumed a value of 86Sr/88Sr of 0.1194 to correct for fractionation.
The final 87Sr/86Sr dataset consists of 190 grass samples from all major physiological provinces in the study area. QGIS 2.10 (Pisa) was used to interpolate and render 87Sr/86Sr as a continuous surface. 87Sr/86Sr data was interpolated using an inverse-distance-weighted algorithm (power = 3; search radius = 0) with no smoothing. An IDW algorithm was preferred over an ordinary kriging method because 87Sr/86Sr of surface sediments throughout most of the study area vary continuously. Unfortunately, this method produces artifacts in areas of discrete variation such as the Ozark uplift and Black Hills. Future studies that seek to understand second-order variability in these areas should take this into account. Surface profiles in well-sampled areas were rendered with the Profile Tool (ver. 3.7.0) plug-in.
Surface Sr isotope values vary significantly across the Midcontinent (Figure 1). The most radiogenic samples were from Precambrian igneous outcrops in the Black Hills and Missouri Ozarks (Harney Peak Granite, SD: 0.7484; Graniteville Granite, MO: 0.7330). Outlying areas of these bedrock-dominated landscapes show lower values consistent with younger carbonate rocks (0.7139–0.7179). Glaciogenic sediments in the upper Midwest containing shield-derived clasts also exhibit high values (0.7162–0.7122). The lowest 87Sr/86Sr values are found in the western part of the study area, where valley alluvium and uplands are dominated by Neogene sources and loess-covered landscapes along the Missouri River and the eastern prairies (0.7079–0.7120). Mississippian and Pennsylvanian limestones in Illinois, Missouri, and eastern Kansas (0.7093–0.7085) and Cretaceous limestone in western Kansas (0.7076) also exhibit low 87Sr/86Sr values. Although not mutually exclusive, these geologic packages represent large-scale Sr isotope variability in surface materials across the Midcontinent.
Sampling resolution in the study area was not high enough to establish sub-regional patterns within larger geologic packages. However, it is possible to illustrate local changes in well-sampled or highly variable areas (Figure 2). Although 87Sr/86Sr profiles constructed across the Sr isoscape show expected changes across older landscapes such as the Ozarks and Black Hills, locally significant variability in 87Sr/86Sr also occurred in other parts of the study area where alluvial valleys show local, negative excursions when compared to surrounding uplands. This is true for landscapes in the Great Plains, where uplands are loess-capped bedrock, as well as glaciated parts of the Midwest, where uplands consist of till and loess (Figure 2).
Ancient landscapes comprised of Precambrian igneous and metamorphic rocks with little alluvial or eolian surface deposits are isotopically heterogeneous and exhibit the highest 87Sr/86Sr values in the study area. Widespread limestones outcropping near the surface in Kansas, Missouri, and Illinois show significantly lower 87Sr/86Sr values (consistent with the seawater curve; McArthur et al., 2001) than crystalline rocks in the Ozark Uplift and Black Hills. Since these rock types are easily eroded, they potentially comprise more of the bedload in regional rivers and streams.
In glaciated areas, glacial till exerts a strong influence on Sr isotope variability. For instance, glaciogenic deposits sourced in the granite-dominated Precambrian Canadian Shield show higher 87Sr/86Sr values than tills with source areas dominated by Cretaceous-aged Dakota sandstone. In these areas, loess and lacustrine sediments are also predominantly (although not entirely) derived from glacial sources. Local-scale variability in the characteristics of surface sediments from glaciated areas is much higher than the sample resolution in this dataset (Figure 1D). Therefore second-order variability (e.g., variability of 87Sr/86Sr within geologic units) is beyond the scope of this study.
The range of 87Sr/86Sr values in this study are similar to previously published data from the region utilizing small or non-migratory animals (Price et al., 2007; Hedman et al., 2009; Slater et al., 2014), sediments, and grasses (Widga et al., 2010). However, the spatial scale of this isoscape is much greater and demonstrates the potential for significant local variability in surface sediments.
Comparisons can be made between this surface dataset and bedrock-derived Sr isoscapes (Figure 3). Bataille and Bowen (2012) built two large-scale, process-driven isoscapes using widely available geological data (available at wateriso.utah.edu/waterisotopes, accessed on 21 April, 2017). The resulting coverages reflect 1) age of bedrock (Beard and Johnson, 2000; Figure 6B in Bataille and Bowen, 2012) and 2) local water (Figure 9A in Bataille and Bowen, 2012). Due to the presence of deep surficial sediments in the Great Plains and Midwest, residual maps of surface 87Sr/86Sr values presented in this study and these bedrock-derived isoscapes show significant differences. These differences are especially pronounced in areas of deep glaciogenic deposits in the upper Midwest where bedrock models overestimate 87Sr/86Sr values by up to 0.0333. The bedrock model more closely aligns with the surface data in unglaciated parts of the Great Plains, but both positive and negative residuals are present in the Midwest, where loess-dominated landscapes reflect the source of the loess, not the underlying bedrock.
These data illustrate the difficulty in constructing regional Sr isoscapes. Although continent-scale changes can be estimated through bedrock derived models, these models cannot presently account for highly variable surface expression of 87Sr/86Sr, especially in areas exposed to significant deposition of loess, alluvium, and glacial till. Provenance and mobility studies in these areas should take this into account.
Sr isotope tracers are increasingly useful for understanding a broad range of ecological and geological questions. In many applications, such as understanding wildlife movement patterns (Hoppe et al., 1999; Balasse et al., 2002; Hoppe, 2004; Hoppe and Koch, 2007; Britton et al., 2009; Britton et al., 2011) or forensic provenance (Beard and Johnson, 2000; Degryse et al., 2012), it is important to establish the maximum degree of variability in bioavailable Sr. We illustrate the structure of 87Sr/86Sr variability across the Midcontinent, encompassing a wide range of surface geology, vegetation, and terrain.
For mobility and provenance studies, bedrock and water models of 87Sr/86Sr distribution are inadequate for the region due to extensive coverage of allochthonous alluvial, eolian, and glaciogenic sediments. These surfaces are complex, requiring direct sampling of surface deposits, or vegetation growing on surface deposits. Unsurprisingly, the surface dataset presented here differs significantly from bedrock-derived models. Although we would expect the models to be more in agreement in areas without significant surface deposition, this is not always the case. Further refinement of all 87Sr/86Sr isoscape models for the Midcontinent region is required.